Neural Occupancy Map Reconstruction for Low-Bitrate Point Clouds
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies face challenges in efficiently compressing dynamic point clouds for distribution while maintaining high quality and reducing bit-rate consumption, which is crucial for applications like virtual reality and autonomous vehicles.
Innovation Solution
A method and apparatus using a neural network to upscale or downscale occupancy maps of volumetric content, leveraging a two-layer-based point cloud encoding and decoding structure, including a base layer and enhancement layer, to optimize compression and quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If the occupancy map is decoded at a lower resolution to reduce bit-rate consumption, then the compression efficiency is improved, but the reconstruction quality deteriorates
Solution Approach 1:
The patent applies parameter changes by using a neural network to transform the occupancy map from a lower resolution representation to a higher quality reconstructed form. The neural network processes the decoded occupancy map and generates enhanced output that maintains or improves reconstruction quality while benefiting from the compression achieved through lower resolution decoding.
2Manufacturing precision
If a neural network is used to upscale the occupancy map, then the reconstruction quality is improved, but the device complexity increases
Solution Approach 1:
The patent replaces traditional mechanical or algorithmic upscaling methods with a neural network-based approach. This substitution enables higher reconstruction quality by leveraging learned patterns from training data, achieving superior results compared to conventional interpolation or upscaling algorithms while managing computational complexity through optimized network architectures.
Data Source
AI summary
At least one embodiment relates to a method and an apparatus for reconstructing an occupancy map comprising occupancy data of a volumetric content, wherein reconstructing the occupancy map comprises: —decoding the occupancy map at a first resolution, —determining a scale factor as a function of the first resolution, —upscaling the occupancy map by the scale factor, using a neural network.